2. Observations and data reduction
The data were recorded during the commissioning of the grism mode of the ESO mid-infrared camera TIMMI (T hermal I nfrared M ultim ode I nstrument) at the 3.6-m telescope in December 1995. Using the spectroscopy mode of this instrument, we obtained 10 µ m spectra of UX Ori . We used three grisms covering the wavelength ranges from 7.8 µ m to 9.4 µ m, 9.3 µ m to 11.2 µ m, and 10.9 µ m to 13.3 µ m, respectively. The spectral resolution is not exactly the same for the different grisms. It varies from 0.026 µ m/pixel for the 8 µ m grism to 0.037 µ m/pixel for the 12 µ m grism. The wavelength calibration was done in the laboratory by scanning black body radiation with a monochromator. A more detailed technical description of TIMMI was given by Käufl (1994) and Käufl et al. (1994).
In the two nights of December 5 and 6 we obtained two complete spectra (that means two sets of three single spectra according to the different grisms) of UX Ori with a signal-to-noise ratio . In the first night we observed comparison spectra of Car before and after the exposures for UX Ori. In the second night we obtained spectra of Car before and of CMa after the exposures for UX Ori at nearly the same airmass. The signal-to-noise ratio of the spectra of the comparison stars was . Typical integration times for one grism including all copping and nodding cycles were 30 min for UX Ori and 5 min in case of a comparison star.
The wavelength calibration was checked by comparing the peaks in the observed atmospheric transmission curve with the positions of the atmospheric absorption lines (especially O3 around 9.8 µ m). However, we cannot completely exclude a remaining wavelength calibration error in the order of one pixel ( 0.031 µ m). This residual error is due to the limited mechanical accuracy to set the aperture wheel of the camera containing the slits to a given position.
Comparing the two standard stars in each of the two nights during the process of data reduction, it turned out that the atmospheric extinction must have varied rapidly during these two nights because the differences between these spectra could not be accounted for by the different airmasses alone. The deviations reach up to 10 per cent at some wavelengths. Because of this fact we reduced the data in two different ways. First, we used the spectra of the comparison stars to derive the atmospheric transmission and to reduce the programme star spectra and, second, we used model calculations of the atmospheric transmission to obtain the spectra for UX Ori.
As one way to reduce our data we derived the atmospheric transmission by comparing our spectra of Car and CMa with their IRAS LRS spectra (IRAS Science Team 1986). Both IRAS spectra are featureless Rayleigh-Jeans-type spectra of class 18 with a . For the reduction of the TIMMI spectra we interpolated the IRAS spectra for the corresponding pixel wavelengths. In the case of CMa we compared the IRAS spectrum also with the calibrated model spectrum of CMa by Cohen et al. (1992). We found a flux ratio of 0.945 of the model spectrum to the LRS spectrum. This could imply that our fluxes of UX Ori that are based on the IRAS spectra are too large by about 5 per cent. In the reduction procedure for the three wavelength intervals (corresponding to the three grisms), we always used the comparison spectrum that had the shortest time difference from the UX Ori exposures. This time difference was at maximum two hours. The angular difference on the sky between UX Ori and the comparison stars was 25 and 50 for Car and CMa, resp. The resulting differences in airmass X were in no case larger than = 0.20. After the reduction the three parts of the spectra fit well together. The mean of the two spectra of UX Ori is presented in the upper panel of Fig. 1. The good agreement can clearly be seen in the overlapping region between the 10 µ m grism and the 12 µ m grism around 11 µ m. The lower panel of Fig. 1 shows the combined effect of atmospheric extinction and sensitivity of the IR-array. The observational error is approximately inversely proportional to the signal. It is obvious that the noise increases considerably towards both atmospheric cut-offs and at the position of the O3 absorption band.
As a second reduction procedure we used the atmospheric transmission model by Wooden (1996, priv. comm.) to remove the effects of the atmosphere. We calculated the atmospheric transmission for the airmass X using the formula
where and are functions given by the model. For fine-tuning we varied the airmass until we got the best fit to the ratio of the IRAS spectra and the observed spectra of Car and CMa. The result is virtually the same as that shown in Fig. 1. However, it turned out that, as in the first procedure, the influence of the atmosphere could not be fully removed by adjusting the transmission model by the airmass at the time of observation alone. During the whole sequence of observations the amount of water vapour and ozone must have changed.
© European Southern Observatory (ESO) 1997
Online publication: April 20, 1998